/
newgpt.py
187 lines (157 loc) · 7.14 KB
/
newgpt.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import logging
import os
import sys
from typing import Dict, List, Optional
from .hf_newgpt import NewGPTConfig, NewGPTForCausalLM
import torch
import torch.nn as nn
from fairseq import checkpoint_utils
from fairseq.models import (
FairseqIncrementalDecoder,
FairseqLanguageModel,
register_model,
register_model_architecture,
)
logger = logging.getLogger(__name__)
@register_model("newgpt")
class NewGPTLanguageModel(FairseqLanguageModel):
def __init__(self, decoder):
super().__init__(decoder)
@staticmethod
def add_args(parser):
"""Add model-specific arguments to the parser."""
# fmt: off
parser.add_argument('--gpt-model-path', default="", help='checkpoint path')
parser.add_argument('--embed-dim', type=int, metavar='N',
help='embedding dimension')
parser.add_argument('--num-attention-heads', type=int, metavar='N',
help='num attention heads')
parser.add_argument('--num-layers', type=int, metavar='N',
help='num layers')
parser.add_argument('--dropout', type=float, metavar='D',
help='dropout probability for all fully connected layers '
'in the embeddings, encoder, and pooler')
parser.add_argument('--attention-dropout', type=float, metavar='D',
help='dropout probability for attention weights')
parser.add_argument('--newgpt-mode', type=str, metavar='N',
help='change the model structure')
parser.add_argument('--result-path', type=str, metavar='N',
help='path to save results',
default="./output/debug.json")
parser.add_argument('--retrieval-layer-index', type=int, metavar='N',
help='The layer index for retrieval',
default="17")
parser.add_argument('--use-external-memory', action="store_true",
help='use external memory or not',
default=False)
# fmt: on
@classmethod
def build_model(cls, args, task):
"""Build a new model instance."""
default_architecture(args)
model = cls(NewGPTDecoder(args, task))
if args.gpt_model_path != "":
state = checkpoint_utils.load_checkpoint_to_cpu(args.gpt_model_path)
model.load_state_dict(state["model"], strict=True, args=args)
return model
class NewGPTDecoder(FairseqIncrementalDecoder):
def __init__(self, args, task):
super().__init__(task.target_dictionary)
config = NewGPTConfig(
vocab_size=len(task.target_dictionary),
# n_ctx=args.max_target_positions,
n_embd=args.embed_dim,
mode=args.newgpt_mode,
max_position_embeddings=args.tokens_per_sample,
n_layer=args.num_layers,
n_head=args.num_attention_heads,
resid_pdrop=args.dropout,
embd_pdrop=args.dropout,
attn_pdrop=args.attention_dropout,
layer_norm_epsilon=1e-6,
retrieval_layer_index=args.retrieval_layer_index,
use_external_memory=getattr(args, "use_external_memory", False),
)
self.model = NewGPTForCausalLM(config)
# set zero embedding for padding symbol
self.pad_idx = task.target_dictionary.pad()
self.model.transformer.wte.weight.data[self.pad_idx].zero_()
def forward(
self,
prev_output_tokens,
src_lengths=None,
incremental_state: Optional[Dict[str, List[torch.Tensor]]] = None,
encoder_out=None,
features_only: Optional[bool] = False,
return_all_hiddens: Optional[bool] = False,
disable_add_index: Optional[bool] = False,
):
features, all_hidden_states, kv_dict, position_encoding = self.extract_features(prev_output_tokens, incremental_state, disable_add_index=disable_add_index)
lm_logits = self.model.lm_head(features)
if return_all_hiddens:
return lm_logits, all_hidden_states, kv_dict, position_encoding
if features_only:
return features
return (lm_logits, None)
def extract_features(
self,
prev_output_tokens,
incremental_state: Optional[Dict[str, List[torch.Tensor]]] = None,
disable_add_index: Optional[bool] = False,
):
if incremental_state:
past = self.get_incremental_state("past")
else:
past = None
# don't attend to padding symbols
attention_mask = prev_output_tokens.ne(self.pad_idx).int()
outputs, kv_dict, position_encoding = self.model.transformer(
input_ids=prev_output_tokens,
past_key_values=past,
attention_mask=attention_mask,
output_hidden_states=True,
disable_add_index=disable_add_index,
)
last_hidden_state = outputs.last_hidden_state
all_hidden_states = outputs.hidden_states
if incremental_state:
self.set_incremental_state(incremental_state, "past", outputs[1])
return last_hidden_state, all_hidden_states, kv_dict, position_encoding
@register_model_architecture("newgpt", "newgpt-small")
def default_architecture(args):
args.embed_dim = getattr(args, "embed_dim", 768)
args.num_attention_heads = getattr(args, "num_attention_heads", 12)
args.num_layers = getattr(args, "num_layers", 12)
args.dropout = getattr(args, "dropout", 0.1)
args.attention_dropout = getattr(args, "attention_dropout", 0.1)
args.tokens_per_sample = getattr(args, "tokens_per_sample", 2048)
args.newgpt_window = getattr(args, "newgpt_window", args.tokens_per_sample)
args.retrieval_layer_index = getattr(args, "retrieval_layer_index", 10)
@register_model_architecture("newgpt", "newgpt-mini")
def newgpt_medium(args):
args.embed_dim = getattr(args, "embed_dim", 128)
args.num_attention_heads = getattr(args, "num_attention_heads", 4)
args.num_layers = getattr(args, "num_layers", 8)
default_architecture(args)
@register_model_architecture("newgpt", "newgpt-medium")
def newgpt_medium(args):
args.embed_dim = getattr(args, "embed_dim", 1024)
args.num_attention_heads = getattr(args, "num_attention_heads", 16)
args.num_layers = getattr(args, "num_layers", 24)
default_architecture(args)
@register_model_architecture("newgpt", "newgpt-large")
def newgpt_large(args):
args.embed_dim = getattr(args, "embed_dim", 1280)
args.num_attention_heads = getattr(args, "num_attention_heads", 20)
args.num_layers = getattr(args, "num_layers", 36)
default_architecture(args)
@register_model_architecture("newgpt", "newgpt-xl")
def newgpt_xl(args):
args.embed_dim = getattr(args, "embed_dim", 1600)
args.num_attention_heads = getattr(args, "num_attention_heads", 25)
args.num_layers = getattr(args, "num_layers", 48)
default_architecture(args)